As more CX teams deploy AI agents, the build-versus-buy question is no longer theoretical. Leaders are feeling real pressure to move beyond pilots and prove impact, but many are also learning that an agent that looks great in a demo can fall apart fast in production.
In this CX Today interview, Dhwani Soni, Global VP of Product Management and Design at 8×8, breaks down why so many organisations want to build their own AI agents instead of buying off-the-shelf tools. Her view is that the appetite comes down to trust and domain fit. Most businesses do not run “average” workflows. They have their own terminology, escalations, integrations, and regulatory constraints, and they need agents that reflect that reality.
She also explains why architecture matters more than most teams expect. When AI is bolted onto an existing platform, teams can inherit integration debt, transcription intermediaries, API latency, and context loss during call transfers. Those technical gaps quickly become customer experience problems, especially when customers have to repeat themselves or the bot cannot hand off cleanly to a human.
“So when something is bolted on, the failure mode is now starting to get understood, right? You add an AI layer on top of a platform and you immediately start inheriting the integration debt. You’ve got transcription intermediaries, API latency, you’ve got context loss, especially when a call transfers.”
For teams still stuck between planning and rollout, Soni shares practical advice: start with a workflow you understand, define KPIs upfront, map exceptions and escalations, and design the human handoff from day one. She also flags where guardrails matter most, including regulated industries and sensitive customer data, where platforms need to enforce compliance while business users define the workflow.
The bigger shift she is seeing now is expansion. As early adopters get comfortable with inbound use cases, they are moving into proactive outbound workflows like appointment conversations, follow-ups on open service requests, and customer education, without treating AI like a “magic wand.”
Do not miss the discussion if you are weighing build vs buy, debating native vs bolt-on AI, or trying to close the pilot-to-production gap. Watch the full interview now, and stay ahead of what is changing in CX AI adoption.
